New method of generating spectrum compatible accelerograms using neural networks

A new method is proposed for generating artificial earthquake accelerograms from response spectra. This method uses the learning capabilities of neural networks to developed the knowledge of the inverse mapping from the response spectra to earthquake accelerogram. In the proposed method the neural networks learn the inverse mapping directly from the actual recorded earthquake accelerograms and their response spectra. A two-stage approach is used. In the first stage, a replicator neural network is used as a data compression tool. The replicator neural network compresses the vector of the discrete Fourier spectra of the accelerograms to vectors of much smaller dimension. In the second stage, a multi-layer feed-forward neural network learns to relate the response spectrum to the compressed Fourier spectrum. A simple example is presented, in which only 30 accelerograms are used to train the two-stage neural networks. This example demonstrates how the method works and shows its potential. © 1998 John Wiley & Sons, Ltd.     

The generation of spectrum compatible accelerograms for the design of nuclear power plants

Based on the main features of the computer program THGE, the paper reviews the techniques which are available for the construction of an accelerogram whose response spectrum matches a design spectrum. First, a sample accelerogram is generated as the product of the stationary random sequence by a deterministic shape function. It is assumed that the spectral properties of the stationary process have been determined in a previous step, in order to lead to the design spectrum as expected maximum responses of a set of single degree of freedom oscillators. The principal properties of the Fast Fourier Transform which is used for this first step are reviewed. The paper then describes the procedures which are available, both in the frequency domain and in the time domain, to improve the agreement between the response spectrum and the target. It also discusses some related issues, such as the response spectrum calculations, the statistical dependence between the three earthquake components, the duration of the time history, the variability of the secondary response to various samples and the generation of an accelerogram whose response spectra envelop a set of design spectra. The point of view adopted is the one of the structural engineer.     

强震加速度记录的数据处理方法

本文用数值积分法,对RDZ1-12-66型自动触发电流计记录式强震仪的幅频响应失真进行校正。并采用高通数字滤波的方法,修正加速度图的零线.为此编制了计算机程序,绘制了修正后的加速度、速度和位移时程曲线.计算了修正前后加速度图的傅氏谱。同时,对这些结果加以讨论。 修正后的加速度图,精确地表示了仪器基本频带在0.09HZ和25HZ之间的绝对地面加速度。      

基于小波变换的拟合规范反应谱多维地震动模拟

本文提出一种基于小波变换的拟合规范反应谱的多维地震动模拟算法。首先将规范反应谱推广到三维相关设计反应谱,然后将已有的三维地震动加速度时间历程曲线分解为一系列不同频段上的地震动分量,调整每一个地震动分量的幅值使其在相应的频率范围内拟合设计反应谱,最后经过调整后的地震动分量进行重构得到更新的地震动时间历程曲线。将该时间历程曲线的反应谱与目标反应谱进行比较,重复该过程直到误差位于特定的范围内。该方法可以保留原始地震动的局部时-频特性,为多维地震动的模拟提供了一种新的方法。     

Prediction of earthquake energy input from smoothed fourier amplitude spectrum

This paper attempts to show analytically that the energy-input spectra of damped SDOF systems and undamped MDOF systems excited by an earthquake motion can be predicted by smoothing the Fourier amplitude spectrum of the base acceleration. The spectral window for smoothing in the frequency domain for a damped SDOF system is identical with the probability density function of the time-variant or instantaneous vibration frequency resulting from non-linear hysteresis. The spectral window for an undamped MDOF system is identical with the set of squared participation factors associated with vibration modes. It was found that the increase in damping factor and the increase in participation of higher modes provide wider spectral windows, resulting in more flattened or unaltered energy-input spectra due to enhanced smoothing effects.     

Quick algorithms for computing either displacement,velocity or acceleration of an oscillator

Accurate and efficient algorithms are derived for computing any desired response quantity of a single-degree-of-freedom linear oscillator subjected to an arbitrary forcing function. These algorithms are based on linear interpolation of the excitation between discrete points sampled uniformly in time. In the special case of the undamped oscillator, the displacement algorithm is equivalent to a previously published algorithm for this situation which was derived using a finite element approach. However, contrary to the claim made by the authors of this latter algorithm, it does not give an exact solution of the equation of undamped motion for an arbitrary forcing function.     

A brief theory and computing of seismic ground rotations for structural analyses

The seismic ground rotations are important with respect to spatial structural models, which are sensitive to the wave propagation. The rotational ground motion can lead to significant increasing of structural response, instability and unusual damages of buildings. Currently, the seismic analyses often take into account the rocking and torsion motions separately using artificial accelerograms. We present an exact analytical method, proposed by Nazarov [15] for computing of three rotational accelerograms simultaneously from given translational records. The method is based on spectral representation in the form of Fourier amplitude spectra of seismic waves, corresponding to the given three-component translational accelerogram. The composition, directions and properties of seismic waves are previously determined in the form of a generalized wave model of ground motion. It is supposed that seismic ground motion can be composed by superposition of P, SV, SH- and surface waves. As an example, the dynamic response analysis of 25-story building is presented. Here recorded (low-frequency) and artificial (high-frequency) accelerograms were used; each of them includes three translational and three rotational components. In this structural analysis, we have clarified primarily conditions under which rotational ground motion should be taken into account. Next, we have calculated three rotational components of seismic ground motion. Then they were taken as additional seismic loads components for further seismic analysis of the building. Note, soil–structure interaction (SSI) is not considered in this study. For computing, we use the special software for structural analyses and accelerogram processing (FEA Software STARK ES and Odyssey software, Eurosoft Co., Russia). It was developed and is used in engineering practice in the Central Research Institute of Building Constructions (TsNIISK, Moscow, Russia).     

Wavelet‐based characterization of design ground motions

With the recent emergence of wavelet‐based procedures for stochastic analyses of linear and non‐linear structural systems subjected to earthquake ground motions, it has become necessary that seismic ground motion processes are characterized through statistical functionals of wavelet coefficients. While direct characterization in terms of earthquake and site parameters may have to wait for a few more years due to the complexity of the problem, this study attempts such characterization through commonly available Fourier and response spectra for design earthquake motions. Two approaches have been proposed for obtaining the spectrum‐compatible wavelet functionals, one for input Fourier spectrum and another for input response spectrum, such that the total number of input data points are 30–35% of those required for a time‐history analysis. The proposed methods provide for simulating ‘desired non‐stationary characteristics’ consistent with those in a recorded accelerogram. Numerical studies have been performed to illustrate the proposed approaches. Further, the wavelet functionals compatible with a USNRC spectrum in the case of 35 recorded motions of similar strong motion durations have been used to obtain the strength reduction factor spectra for elasto‐plastic oscillators and to show that about ±20% variation may be assumed from mean to 5 and 95% confidence levels due to uncertainty in the non‐stationary characteristics of the ground motion process. Copyright © 2002 John Wiley & Sons, Ltd.     

Effects of soil–structure interaction on seismic base isolation

The effects of soil–structure interaction on the performance of a nonlinear seismic base isolation system for a simple elastic structure are examined. The steady-state response of the system to harmonic excitation is obtained by use of the equivalent linearization method. Simple analytical expressions for the deformation of the base isolation system and of the superstructure at resonance are obtained in terms of an effective replacement oscillator characterized by amplitude-dependent frequency, damping ratio, and excitation. Numerical results suggest that the seismic response of a structure resting on an inelastic base isolation system may be larger when the flexibility of the soil is considered than the corresponding response obtained by ignoring the effects of soil–structure interaction. It is shown that, in the undamped case and in the absence of soil–structure interaction effects, a critical harmonic excitation exists beyond which the steady-state resonant response of the isolators and structure become unbounded.     

Earthquake response analysis considering non-proportional damping

Non-proportional damping may be defined as a form of linear viscous damping which introduces coupling between the undamped modal co-ordinate equations of motion. The standard mode superposition method of earthquake response analysis therefore cannot be employed with non-proportionally damped structures. In this paper, several methods for analysing the dynamic response of non-proportional damped structures are outlined. It is concluded that the most efficient procedure is to express the response in terms of a truncated set of undamped modal coordinates and to integrate directly the resulting coupled equations. The effectiveness of the method is demonstrated by a numerical example.     

力平衡式加速度计仪器响应误差的校正方法

本文通过对数字强震仪系统使用的力平衡加速度计的仪器响应误差进行分析,提出了一个校正仪器响应失真的处理方法(微分-微分方法),即对未校正记录做低通滤波后运用近似理想微分器做两次微分得到校正加速度记录。针对大亚湾核电站地震仪表系统(K IS)给出了仪器校正示例,并编制了相应计算软件。该方法适用于我国强震动台网数字强震仪获取的加速度记录的仪器响应误差校正处理。     

The use of optimization to construct critical accelerograms for given structures and sites

In this paper a methodology has been presented for constructing the most critical accelerogram from among a be class of candidate accelerograms for a given site and structure. This most critical accelerogram could be used to assess seismic resistance of a structure with a high level of confidence. Specifically, the method superimposes accelerograms recorded at similar sites to create the candidate accelerograms, then uses optimization and approximation techniques find the most critical accelerogram. The most critical accelerogram is defined as the one which maximizes damage is structure, as computed by non-linear dynamic structural analysis, as well as satisfies constraints on ground parameters to ensure credibility. The damage has been defined as cumulative inelastic energy dissipation or sure of interstorey drifts. The method is applied to ten examples in the paper.     

Wavelet-based generation of spectrum-compatible time-histories

This paper deals with the well-known problem of generating spectrum-compatible synthetic accelerograms for the linear and non-linear time-history analyses of structural systems. A wavelet-based procedure has been used to decompose a recorded accelerogram into a desired number of time-histories with non-overlapping frequency contents, and then each of the time-histories has been suitably scaled for matching of the response spectrum of the revised accelerogram with a specified design spectrum. The key idea behind this iterative procedure is to modify a recorded accelerogram such that the temporal variations in its frequency content are retained in the synthesized accelerogram. The proposed procedure has been illustrated by modifying five recorded accelerograms of widely different characteristics such that those are compatible with the same USNRC design spectrum.     

Aseismic design implications of near-fault san fernando earthquake records

Near-fault records of the 1971 San Fernando earthquake contain severe, long duration acceleration pulses which result in unusually large ground velocity increments. A review of these records along with the results of available theoretical studies of near-fault ground motions indicates that such acceleration pulses may be characteristic of near-fault sites in general. The results of an analytical study of a building severely damaged during the San Fernando earthquake indicate that such severe, long duration acceleration pulses were the cause of the main features of the observed structural damage. The implications of such pulses on current aseismic design methods, particularly those used to establish design earthquakes, are examined for buildings located near potential earthquake faults. Analytical studies of the non-linear dynamic response of single and multiple degree-of-freedom systems to several near-fault records, as well as to a more standard accelerogram, indicate that at near-fault sites: (a) very large displacement ductilities may result for current levels of code design forces; (b) smoothed elastic design response spectra should reflect the larger ground velocities that may occur; and (c) peak inelastic response cannot reliably be inferred from elastic response predictions.     

Damped vibration mode superposition method for dynamic response analysis

In the dynamic response analysis of extremely complex structural systems in which the damping characteristics of each element are independent, the damping matrix is not always diagonalized by the use of undamped free vibration mode shapes. In the present paper, a mode-superposition method by the use of damped free vibration mode shapes is developed for such structural systems. It is also shown that the Fast Fourier Transform (FFT) procedures, that are available for the dynamic response analysis of linear structural systems, are used effectively in this mode-superposition method with good accuracy.     

近场强震动记录基线校正方法的振动台实验研究

为了获取近场永久位移,通常采用基线校正方法,对近场加速度记录进行基线校正并积分得到永久位移值,但这一结果主观性较强,其可靠性也往往缺乏验证。为了解决这一问题,本文提出了一种能产生包含永久位移振动过程的振动台实验方案,采用振动台加滑动机构的方法,模拟记录到永久位移台站测点的真实振动情况;在实验中分别采用加速度计、摄影测量方法分别直接得到加速度和位移时程,对加速度时程进行基线校正并积分得到位移时程,将其与直接获得的位移时程进行对比,以验证采用基线校正方法的有效性。实验结果表明,在实验室条件下采用现有的基线校正方法校正后,通过积分能得到可以接受的位移时程。     

A subspace modal superposition method for non-classically damped systems

For structures with non-proportional damping, complex eigenvectors or mode shapes must be used in order to decoe the equations of motion. The resulting equations can then be solved in a systematic way. The necessity of solvie complex eigenvalue problem of a large system remains an obstacle for the practical application of the method. This stres utilizes the fact that in practice only a small number of the complex modes are needed. Therefore, these complex modes be approximated by a linear combination of a small number of the undamped modes, which can be obtained by established methods with less cost. An additional eigenvalue problem is then solved in a subspace with a much sm dimension to provide the best combination coefficient for each complex mode. The method of solution for the decoue equations is then carried over, using the approximate complex modes expressed in undamped mode shapes, to resue simple formulas for the time- and frequency-domain solution. Thus, an efficient modal superposition method is develoe for non-proportionally damped systems. The accuracy of this approximate method is studied through an example. Comparing the frequency response result using the approximate method with that using the exact complex modes, found that the error is negligible.     

Effect of causal and acausal filters on elastic and inelastic response spectra

With increasing interest in displacement spectra and long‐period motions, it is important to check the sensitivity of both elastic and inelastic response spectra to the filtering that is often necessary to remove long period artifacts, even from many modern digital recordings. Using two records of very different character from the M=7.1, 1999 Hector Mine, California, earthquake, we find that the response spectra can be sensitive to the corner periods used in causal filtering, even for oscillator periods much less than the filter corner periods. The effect is most pronounced for inelastic response spectra, where the ratio of response spectra computed from accelerations filtered at 25 and 200 sec can be close to a factor of 2 for oscillator periods less than 5 sec. Published in 2003 by John Wiley Sons, Ltd.     

Quantification of ground-motion parameters and response spectra in the near-fault region

This study focuses on the characteristics of near-fault ground motions in the forward-direction and structural response associated with them. These ground motions are narrow-banded in nature and are characterized by a predominant period at which structures excited by them are severely affected. In this work, predominant period is defined as the undamped natural period of a single-degree-of-freedom (SDOF) oscillator at which its 5% damped linear elastic pseudo-spectral velocity (PSV) contains a clear and dominant peak. It is found that a linear relationship exists between predominant period and seismic moment. An empirical equation describing this relationship is presented by using a large set of accelerograms. Attenuation equations are developed to estimate peak ground velocity (PGV) as a function of earthquake magnitude and source-to-site distance. In addition, a predictive equation for spectral shapes of PSV (i.e., PSV normalized by PGV) is presented as a continuous function of the undamped natural period of SDOF oscillators. The model is independent of PGV, and can be used in conjunction with any available PGV attenuation relation applicable to near-fault ground motion exhibiting forward-directivity effects. Furthermore, viscous damping of the SDOF is included in the model as a continuous parameter, eliminating the use of so-called damping correction factors. Finally, simple equations relating force reduction factors and displacement ductility of elasto-plastic SDOF systems are presented.     

Generation of accelerograms compatible with design specifications using information theory

This paper deals with the generation of seismic accelerograms which are compatible with a given response spectrum and other design specifications. The time sampling of the stochastic accelerogram yields a time series represented by a random vector in high dimension. The probability density function of this random vector is constructed using the maximum entropy (MaxEnt) principle under constraints defined by the available information (design specifications). In this paper, an adapted algorithm is proposed to identify the Lagrange multipliers introduced in the MaxEnt principle to take into account the constraints. This algorithm is based on (1) the minimization of an appropriate convex functional and (2) the construction of the probability distribution defined as the invariant measure of an Itô stochastic differential equation in order to estimate the integrals in high dimension of the problem. The constraints related to a seismic accelerogram are developed explicitly. This methodology is validated through an application for which the available information is related to the variance of each component of the random vector representing the accelerogram, statistics on the response spectrum, on the peak ground acceleration, on the cumulative absolute velocity and on the end-values for the velocity and for the displacement.